基于单幅正面照和统计模型的三维人脸重建方法研究
|
摘要
为提升单张图片三维重建的效率、提高重建的精准度,提出了一种基于单张图片和统计模型的三维人脸重建技术。首先基于智能行为理解组(intelligent behaviour understanding group,IBUG),采用监督下降算法(SDM)进行训练来建立二维人脸特征点参数模型;再应用该参数模型对给定图片进行特征点提取;然后基于北京工业大学3D数据库(BJUT-3D)进行三维人脸统计模型的构建;最后,以三维人脸统计模型为基础,使用学习因子自适应梯度下降法对能量函数进行迭代优化,得到统计模型参数化向量,使用该向量来调整统计模型,实现与给定图片相匹配的三维人脸模型构建。实验证明,与现有的方法相比,本文方法重建出的三维人脸模型具有更高的精确度和自适应度。
3D face model reconstruction based on a single frontal photo is widely used in practical applications. In order to improve the efficiency and accuracy of 3D reconstruction based on single frontal image, this work proposes an effective approach to reconstruct a 3D face model based on a single image and a statistical model. First, we train with a SDM algorithm to establish a two-dimensional human face feature point parameter model based on the IBUG two-dimensional face database. The parameter model is used to extract feature points from a given image. A 3D face statistics model is then constructed based on the BJUT-3D database. In order to determine the parametric vector of the statistical model, we optimize the energy function with a learning factor adaptive gradient descent method iteratively based on the 3D face statistics model. Finally, we adjust the statistical model to reconstruct the 3D face model of the given image. The experimental results show that our face reconstruction method has higher accuracy and adaptability than representative commonly used algorithms.
3D face model reconstruction based on a single frontal photo is widely used in practical applications. In order to improve the efficiency and accuracy of 3D reconstruction based on single frontal image, this work proposes an effective approach to reconstruct a 3D face model based on a single image and a statistical model. First, we train with a SDM algorithm to establish a two-dimensional human face feature point parameter model based on the IBUG two-dimensional face database. The parameter model is used to extract feature points from a given image. A 3D face statistics model is then constructed based on the BJUT-3D database. In order to determine the parametric vector of the statistical model, we optimize the energy function with a learning factor adaptive gradient descent method iteratively based on the 3D face statistics model. Finally, we adjust the statistical model to reconstruct the 3D face model of the given image. The experimental results show that our face reconstruction method has higher accuracy and adaptability than representative commonly used algorithms.
引文
[1] 李想. 基于单张正面图象的三维人脸建模研究[D]. 苏州: 苏州大学, 2016. LI X. Research on 3D face modeling based on single front image [D]. Suzhou: Soochow University, 2016. (in Chinese)
[2] CHU C H, WANG I J, WANG J B, et al. 3D parametric human face modeling for personalized product design[J]. Advanced Engineering Informatics, 2017, 32: 202- 223.
[3] IZADI S, KIM D, HILLIGES O, et al. Kinect fusion: real-time 3D reconstruction and interaction using a moving depth camera[C]//ACM Symposium on User Interface Software & Technology. Santa Barbara, 2011: 559- 568.
[4] TRAN A T, HASSNER T, MASI I, et al. Regressing robust and discriminative 3D morphable models with a very deep neural network[C]//IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, 2017: 1493- 1502.
[5] LI F M, CHEN K, LIU X H. 3D face reconstruction based on convolutional neural network[C]//International Conference on Intelligent Computation Technology and Automation. Changsha, 2017: 71- 74.
[6] BLANZ V, VETTER T. A morphable model for the synthesis of 3D faces[C]//Proceedings of the 26th Annual Conference on Computer Graphics and Interactive Techniques. Los Angeles, 1999: 187- 194.
[7] XIONG X, TORRE F D L. Global supervised descent method[C]//IEEE Conference on Computer Vision and Pattern Recognition. Boston, 2015: 2664- 2673.
[8] YANG H, JIA X H, PATRAS I, et al. Random subspace supervised descent method for regression problems in computer vision[J]. IEEE Signal Processing Letters, 2015, 22(10): 1816- 1820.
[9] HUBER P, FENG Z H , CHRISTMAS W, et al. Fitting 3D morphable models using local features[C]//IEEE International Conference on Image Processing. Quebec, 2015.
[10] HU G S, YAN F, KITTLER J, et al. Efficient 3D morphable face model fitting[J]. Pattern Recognition, 2017, 67: 366- 379.
[11] MOEINI A, MOEINI H, FAEZ K. Pose-invariant facial expression recognition based on 3D face reconstruction and synthesis from a single 2D image[C]//International Conference on Pattern Recognition. Stockholm, 2014: 1746- 1751.
[12] CHEN Z C. Sensorcam cameras and image registration[J]. Proceedings of the National Academy of Sciences of the United States of America, 2014, 104(27): 11483.
[13] XIONG X, TORRE F D L. Supervised descent method and its applications to face alignment[C]//IEEE Conference on Computer Vision and Pattern Recognition. Portland, 2013: 532- 539.
[14] AMERINI I, BARNI M, CALDELLI R, et al. Counter-forensics of SIFT-based copy-move detection by means of keypoint classification[J]. Eurasip Journal on Image & Video Processing, 2013(1): 1- 17.
[15] 许磊, 黎智辉, 王俊娟, 等. 高精度三维人脸图像数据库[J]. 刑事技术, 2015, 40(2): 94- 97. XU L, LI Z H, WANG J J, et al. High-precision 3D face image database[J]. Criminal Technology, 2015, 40(2): 94- 97. (in Chinese)
[16] SANG G L, LI J, ZHAO Q J. Pose-invariant face recognition via RGB-D images[J]. Computational Intelligence and Neuroscience, 2016(3): 1- 9.
[2] CHU C H, WANG I J, WANG J B, et al. 3D parametric human face modeling for personalized product design[J]. Advanced Engineering Informatics, 2017, 32: 202- 223.
[3] IZADI S, KIM D, HILLIGES O, et al. Kinect fusion: real-time 3D reconstruction and interaction using a moving depth camera[C]//ACM Symposium on User Interface Software & Technology. Santa Barbara, 2011: 559- 568.
[4] TRAN A T, HASSNER T, MASI I, et al. Regressing robust and discriminative 3D morphable models with a very deep neural network[C]//IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, 2017: 1493- 1502.
[5] LI F M, CHEN K, LIU X H. 3D face reconstruction based on convolutional neural network[C]//International Conference on Intelligent Computation Technology and Automation. Changsha, 2017: 71- 74.
[6] BLANZ V, VETTER T. A morphable model for the synthesis of 3D faces[C]//Proceedings of the 26th Annual Conference on Computer Graphics and Interactive Techniques. Los Angeles, 1999: 187- 194.
[7] XIONG X, TORRE F D L. Global supervised descent method[C]//IEEE Conference on Computer Vision and Pattern Recognition. Boston, 2015: 2664- 2673.
[8] YANG H, JIA X H, PATRAS I, et al. Random subspace supervised descent method for regression problems in computer vision[J]. IEEE Signal Processing Letters, 2015, 22(10): 1816- 1820.
[9] HUBER P, FENG Z H , CHRISTMAS W, et al. Fitting 3D morphable models using local features[C]//IEEE International Conference on Image Processing. Quebec, 2015.
[10] HU G S, YAN F, KITTLER J, et al. Efficient 3D morphable face model fitting[J]. Pattern Recognition, 2017, 67: 366- 379.
[11] MOEINI A, MOEINI H, FAEZ K. Pose-invariant facial expression recognition based on 3D face reconstruction and synthesis from a single 2D image[C]//International Conference on Pattern Recognition. Stockholm, 2014: 1746- 1751.
[12] CHEN Z C. Sensorcam cameras and image registration[J]. Proceedings of the National Academy of Sciences of the United States of America, 2014, 104(27): 11483.
[13] XIONG X, TORRE F D L. Supervised descent method and its applications to face alignment[C]//IEEE Conference on Computer Vision and Pattern Recognition. Portland, 2013: 532- 539.
[14] AMERINI I, BARNI M, CALDELLI R, et al. Counter-forensics of SIFT-based copy-move detection by means of keypoint classification[J]. Eurasip Journal on Image & Video Processing, 2013(1): 1- 17.
[15] 许磊, 黎智辉, 王俊娟, 等. 高精度三维人脸图像数据库[J]. 刑事技术, 2015, 40(2): 94- 97. XU L, LI Z H, WANG J J, et al. High-precision 3D face image database[J]. Criminal Technology, 2015, 40(2): 94- 97. (in Chinese)
[16] SANG G L, LI J, ZHAO Q J. Pose-invariant face recognition via RGB-D images[J]. Computational Intelligence and Neuroscience, 2016(3): 1- 9.